Not long ago, I was tutoring a Chinese student from overseas in Mandarin. I wanted to leave him with something, so I tried making a few things with Gemini 2.5. They were very simple, but at least I had finally taken the first step toward putting the idea into practice. 🥹
First, I wanted to add pinyin, highlighted keywords, pronunciation, and example sentences to all the class materials. But adding pinyin everywhere made the interface look unattractive, so I removed it. Pinyin and explanations for the key words and phrases were enough.
Then I turned the important vocabulary from class into flashcards, a tile-matching game, and quizzes... My stereotypically Chinese way of thinking kicked in 😂
(For the Chinese vocabulary in the HTML files above, I used a local audio library so each word can be clicked to hear its pronunciation.)
Once an idea has the beginnings of a real form, everything afterward becomes easier. Let your imagination run and put AI to work! 🫶
Then Gemini 3 came out recently. I ran off in the middle of the night on launch day to make what I wanted. Yet I still do not understand why Gemini can read the audio file I upload but cannot transcribe it for me. That is the only piece missing [crying R]. Maybe it is an official restriction?
P.S. Although I did not continue building products related to this later, it still records an exploration I made at the time. Every exploration since then has benefited from it.